The analytic depth of cognitive neuroscience is, in many ways, a curse. Those aspects of high-level cognition most relevant to real-world applications are the least understood at a neurobiological level, and those mechanisms that are well-understood neurobiologically are too simple to inform real-world practices.
The explanatory gaps between these levels of analysis is a result of hyper-reductionism in science, itself rooted in a lasting preference (reverence?) for the simplistic and “parsimonious.” But natural phenomena, like the emergence of behavior from the brain, are ultimately more complex than these reductionistic methods presume, or can account for. Below I review a second domain of cognitive neuroscience in which the shortcomings of reductionism are all too clear, and describe how an integrative reconstructionist approach can help repair the science of higher-level cognition.
Humans have the remarkable capacity to recognize and use abstract symbols, putatively reflected in behaviors as rich and diverse as representational play, the use of words, metaphors, numbers, pictoral representation, and the solution of algebraic equations.
However, there are clear reasons to believe that symbol use is not a unitary capacity, but rather highly graded and heterogenous.
For example, children acquire an understanding of pictoral representation gradually, showing declining manual exploration of photographs and eventually adult-like regard for photos by 18-months. However, they do not extend this supposed capacity for “symbol use” to the solution of algebraic equations for many years (and even then do so only with great difficulty).
This disunity in symbol use is actually mirrored by deep theoretical disagreements about the nature of symbols. Some suggest that humans only utilize six types of symbols whereas others seem to suggest that almost anything can be a symbol (“something that someone intends to refer to something …” from DeLoache, 2004). Thus symbol use seems too complex to be an object of direct scientific study, at least at these theoretical and behavioral levels of analysis.
In contrast, neurobiological investigations of symbol use can seem too simplistic: “symbols” are alternately considered to be the arbitrary words or pictures paired with a particular task, the neuronal responses specific to a task but invariant to its stimuli, and the neural representation of categories like “color.”
Although interesting and important, some might claim these findings are irrelevant to the larger questions surrounding the behavioral correlates of symbol use, and associated possibilities for applying this understanding. Others might counter that understanding the role of symbols in applied domains requires reducing the construct to these simpler, more tractable, and unfortunately less rich subcomponents.
And here is the crucial problem with hyper-reductionism (what I previously and incorrectly referred to as Occam’s Razor). The preference for theories with relatively fewer assumptions (among all those other theories which are identical in explanatory power) ultimately leads us towards highly simplistic models of nature which apply only to one level of analysis, if they apply at all.
Here again, the reconstructionist approach can help to resolve the dilemma between intractable complexity on one hand and over-simplification on the other. For example, it may be that the mechanisms guiding the representation of abstract categories like “color” are the same in principle to those guiding more abstract symbols as in metaphors. Similarly, the neuronal responses to abstract “nonmatch” tasks may be similar in nature to those underlying algebraic rules. We can only know by attempting to reconstruct such high-level behaviors from these lower-level computational models and neurobiological phenomena.
The cognitive neuroscience of high-level cognition is stretched thin with tension between the too-complex and the too-simple, as demonstrated above in the sample domain of symbol use. In seeking to explain complex behavior, neuroscience has isolated myriad neurobiological mechanisms which seem too simple and narrow to explain something like the use of symbols. (And I imagine these tensions are not unique to brain science).
Reconstructionist approaches may provide a solution to the dilemma between the study of intractable complexity on the one hand and the methods of narrow reductionism on the other. Computational modeling can specify how these reduced components give rise to large-scale behavior, and how to actionably and mechanistically understand such behavior in clinical or applied settings.
And this is the blessing of analytic depth – reconstructionist and integrative approaches can pull from multiple levels of analysis to provide a more coherent and actionable understanding of natural phenomena than the “infinite reduction” method dominating mainstream science.